Center for Humane Technology

Theory of Change

CHT articulates a two-step feedback loop: create clarity about the misaligned incentives driving tech design, then transform those incentives through policy, litigation, and public awareness campaigns.

From their website: "We demystify the incentives at play in the tech ecosystem in order to drive new levels of awareness and agency... After achieving clarity, we identify and implement targeted interventions to produce systems-level change."

In practice, this has meant: (1) mass-scale media productions that reach mainstream audiences (The Social Dilemma reached 100M+ viewers; The AI Dilemma has 2.6M+ YouTube views); (2) Congressional testimony and policymaker briefings; (3) expert consulting on litigation against tech companies (Meta/Google trials 2026); and (4) a sustained podcast ("Your Undivided Attention," 27.5M+ downloads, 153+ episodes).

Since 2023, CHT has pivoted from social media harms to AI existential risk. Harris's TED2025 talk frames this as seeking a "narrow path" between decentralized AI chaos and centralized AI dystopia, where "power is matched with responsibility at every level." The specific mechanisms proposed include: restricting AI companions for minors, product liability for AI harms (AI LEAD Act), whistleblower protections for AI insiders, and court-ordered design changes via litigation.

What They Do

Media and awareness. CHT's most distinctive capability. The Social Dilemma (2020) is the most-watched AI-safety-adjacent media product ever created -- 100M+ viewers, 2 Emmys, cited by state attorneys general in subsequent lawsuits. The AI Dilemma (2023) brought AI risk concerns to a mainstream audience. The AI Doc (2026, Sundance) features Harris alongside AI lab CEOs. The podcast has hosted Frances Haugen (first long-form post-whistleblower interview), Yuval Harari, Audrey Tang, and Jonathan Haidt.

Litigation and policy. CHT serves as expert consultant in Character.AI lawsuits. Aza Raskin testified as a fact witness in the New Mexico Meta trial (2026), where Meta was found "maximally guilty" ($375M damages, with injunctive relief pending -- potentially the most impactful outcome, as courts could mandate product design changes). CHT endorsed the bipartisan AI LEAD Act creating product liability for AI. Harris has testified before Congress three times (2019, 2020, 2021).

Elite access. CHT has been at Davos (Human Change House) for three consecutive years. Raskin co-chairs the WEF Global AI Counsel. CHT claims to have briefed 500+ public officials. Harris has appeared on Joe Rogan (twice), Tim Ferriss, 60 Minutes, TED (twice), and every major news outlet.

Education. "Foundations of Humane Technology" course: 17,000+ participants from 130+ countries. The course targets product designers and technologists.

Not a research organization. CHT produces no technical AI safety research -- no alignment work, no interpretability, no evaluations, no benchmarks. Its output is entirely in the awareness/advocacy/policy space.

Key People

Tristan Harris (Co-Founder, President): Former Google Design Ethicist, Stanford Persuasive Tech Lab (BJ Fogg). Star of The Social Dilemma. Time100 Next 2021. The primary public face of CHT. Compensation: $249,891 (FY2024). Speaking fees reportedly $100K-$200K+ per appearance (separate from CHT salary).

Aza Raskin (Co-Founder): Inventor of infinite scroll, son of Jef Raskin (inventor of Macintosh). Co-founded Earth Species Project. Co-chairs WEF Global AI Counsel. Compensation: $210,346 as advisor.

Team is ~20 staff. Three Executive Directors in six years (Fernando, Barcay, Guirado), with Harris as the constant. Board includes leaders from Change.org, Kiva, NationBuilder, and Anaconda -- social enterprise backgrounds, no technical AI safety expertise.

Money and Incentives

Revenue is highly volatile. FY2018: $0.9M. FY2020 (Social Dilemma): $5.2M. FY2023 (peak): $9.5M. FY2024: $3.6M. Expenses nearly doubled to $6.3M in FY2024, creating a major deficit. At current burn rate, CHT has roughly one year of runway from ~$6M in net assets.

Funded by mainstream progressive foundations, not the AI safety ecosystem. Named donors: Ford Foundation ($500K total), Packard Foundation ($600K), Knight Foundation, Open Society Foundations, Omidyar Network ($300K), Craig Newmark Philanthropies, Robert Wood Johnson Foundation, Pritzker entities, Susan Crown Exchange. One AI safety-adjacent donor: Future of Life Institute ($500K in 2023). Zero funding from Coefficient Giving/Open Philanthropy, SFF, or EA-aligned sources.

No AI lab ties. No compute credits, no lab funding, no revenue-sharing arrangements. This gives CHT financial independence that few AI safety orgs have.

Speaking fee economy. Harris's personal speaking fees ($100K-$200K+ per engagement) potentially exceed his CHT salary. This creates a personal incentive to maintain alarm-level rhetoric that commands high speaker fees, even if it doesn't perfectly track with epistemic accuracy.

Financial health is concerning. The FY2024 deficit ($3.6M revenue vs. $6.3M expenses) suggests CHT may have been spending down reserves from the FY2023 windfall. The "contributions" line on 990s is consistently much higher than revenue (e.g., $23M vs. $9.5M in FY2023), suggesting multi-year restricted grants that distort the annual picture.

What Others Say

Structural critique (LibrarianShipwreck, 2018 and 2021): CHT "rushes in to ensure that this space [for tech criticism] is occupied by those who maintain close ties to the tech world." Harris tells Congress "they're not evil" and "they meant well," providing cover that the tech companies themselves could never deliver. CHT reduces citizens to consumers and elevates technologists as saviors. "Humane technology" means social media that is 25% less intrusive, not a fundamentally different relationship with technology.

Evidence critique (Nir Eyal, 2020): The Social Dilemma conflates engagement with addiction, lacks causal evidence, uses the illusion of powerlessness as dramatic technique, and offers no actionable solutions. "These tactics are good, but they're not THAT good."

Framing critique (Cyborgology, 2018): CHT frames the problem as "attention" when the deeper issue is data extraction and surveillance capitalism. CHT doesn't use phrases like "big data" or "personal data" and ignores existing digital rights communities.

Philosophical critique (The Frailest Thing, 2018): "Humane technology" will flounder because there is no public consensus about human flourishing. CHT shows "little awareness of or interest in a longstanding history of tech criticism" -- Ellul, Mumford, Illich, Postman all said similar things decades earlier.

Financial incentives critique (Chuck Russell, 2025): Pattern of pivoting to trending topics, $100K-$200K speaking fees creating personal incentives toward alarm, some claims rely on anonymous hearsay. Harris reportedly admitted AI existential risk seemed like "science fiction" until ChatGPT.

Defenders note: CHT reaches audiences no other AI safety org can touch. 100M viewers is unprecedented. The Meta trial verdicts and Character.AI lawsuits represent concrete policy outcomes. CHT's independence from both AI labs and the EA ecosystem gives it credibility with mainstream policymakers.

What's Absent

CHT has zero engagement with the EA/rationalist AI safety community -- no forum presence, no collaborations, no shared funders. This is the most striking feature of CHT's positioning: it operates in a completely separate intellectual ecosystem from technical AI safety.

No technical work. CHT produces no alignment research, interpretability papers, evaluations, or benchmarks. No technical AI staff. No engagement with the substance of AI alignment.

No formal impact measurement. CHT claims 47 states passed child safety laws "since" The Social Dilemma, but provides no methodology for causal attribution. No annual report exists.

No whistleblower policy despite publicly advocating for whistleblower protections at AI companies.

No self-correction. The AI Dilemma's claim that "2024 will be the last human election" has not been retracted or contextualized. The "50% of AI researchers" extinction probability statistic is presented without survey methodology caveats.

No engagement with open-source AI or supply chain ethics. "Humane technology" means humane for users in wealthy countries, not for miners, factory workers, or communities near e-waste dumps.